List your company

Legal Knowledge Evaluation on CaseHOLD and MMLU-L

79CaseHOLD ScoreGemini-2.5-Pro (25-06-05)
37.29648.12358.9569.777Jan 27, 2026

Evaluation results

MethodLinks
Gemini-2.5-Pro (25-06-05)Params=N/A, Label=false, Knowledge=false2026.017992.685.8
Baseline-w/-labelParams=7B, Label=true, Knowledge=false2026.0176.978.577.7
GPT-4.1 (25-04-14)Params=N/A, Label=false, Knowledge=false2026.0176.491.784.1
Baseline-w/-labelParams=3B, Label=true, Knowledge=false2026.0175.669.472.5
DeepSeek-R1Params=671B, Label=false, Knowledge=false2026.0174.691.783.2
RLAIF (GRPO)Params=7B, Label=false, Knowledge=false2026.0174.47976.7
DGRC (Ours)Params=7B, Label=false, Knowledge=false2026.0174.379.376.8
Baseline-w/-labelParams=1.5B, Label=true, Knowledge=false2026.0173.865.369.6
Baseline-w/o-labelParams=7B, Label=false, Knowledge=false2026.0173.678.576.1
DGRC (Ours)Params=3B, Label=false, Knowledge=false2026.0173.172.772.9
RLAIF (GRPO)Params=3B, Label=false, Knowledge=false2026.017370.871.9
Baseline-w/o-labelParams=3B, Label=false, Knowledge=false2026.0172.669.471
DGRC (Ours)Params=1.5B, Label=false, Knowledge=false2026.0171.567.869.7
Qwen2.5-InstructParams=72B, Label=false, Knowledge=false2026.0171.486.879.1
RLAIF (GRPO)Params=1.5B, Label=false, Knowledge=false2026.0171.264.567.9
Baseline-w/o-labelParams=1.5B, Label=false, Knowledge=false2026.0170.362.866.6
Llama3.1-InstructParams=70B, Label=false, Knowledge=false2026.0169.987.678.8
Qwen2.5-InstructParams=32B, Label=false, Knowledge=false2026.0168.284.376.3
Qwen2.5-InstructParams=7B, Label=false, Knowledge=false2026.0163.975.269.6
LegalHalParams=8B, Label=false, Knowledge=true2026.0163.773.668.7
MegaScienceParams=7B, Label=false, Knowledge=true2026.0160.58170.8
Llama3.1-InstructParams=8B, Label=false, Knowledge=false2026.0157.776.967.3
MegaScienceParams=3B, Label=false, Knowledge=true2026.0152.466.159.3
Qwen2.5-InstructParams=1.5B, Label=false, Knowledge=false2026.0144.161.252.7
MegaScienceParams=1.5B, Label=false, Knowledge=true2026.0140.963.652.3
Qwen2.5-InstructParams=3B, Label=false, Knowledge=false2026.0138.961.250.1